Skip to Main Content
Fault classification in electric power system is vital for secure operation of power systems. It has to be accurate to facilitate quick repair of the system, improve system availability and reduce operating costs due to mal-operation of relay. Artificial neural networks (ANNs) can be an effective technique to help to predict the fault, when it is provided with characteristics of fault currents and the corresponding past decisions as outputs. This paper describes the use of particle swarm optimisation (PSO) for an effective training of ANN and the application of wavelet transforms for predicting the type of fault. Through wavelet analysis, faults are decomposed into a series of wavelet components, each of which is a time-domain signal that covers a specific octave frequency band. The parameters selected for fault classification are the detailed coefficients of all the phase current signals, measured at the sending end of a transmission line. The information is then fed into ANN for classifying the faults. The proposed PSO-based multi-layer perceptron neural network gives 99.91% fault classification accuracy. Moreover, it is capable of producing fast and more accurate results compared with the back-propagation ANN. Extensive simulation studies were carried out and a set of results taken from the simulation studies are presented in this paper. The proposed technique when combined with a wide-area monitoring system would be an effective tool for detecting and identifying the faults in any part of the system.